The essential facility concept has been used extensively in antitrust cases in several jurisdictions, including Brazil.7 The Essential Facilities Doctrine (EFD) was originally derived from a 1912 case (United States v. Terminal Railroad Ass'n of St. Louis).  In this case, the US Supreme Court ruled that a consortium of railroad facilities that carried freight traffic and passengers across the Mississippi River in St. Louis must comply with the following conditions: (i) allowing other railroads into the consortium; (ii) charging non- consortium railroads infrastructure usage fees that would place them on nearly equal footing as the consortium members.8  Lower courts in the USA subsequently established the doctrine that, under antitrust law, requires a monopolist firm controlling a facility, which cannot be replicated (due to technical or financial reasons) and is essential for competitors, to provide reasonable access to this facility for competitors.

Hence, two crucial questions arise: how to demonstrate the "essentiality" of a facility and how to establish mechanisms that ensure fair access to this facility for competitors? The EFD would be aptly applied to a subset of practices categorized as refusals to deal with a competitor. Graeff (2020) argues that the EFD is applicable in cases where new customers are denied service, while a refusal to deal refers to supply “there is a pre-existing course of dealing between the dominant firm and the access seeker" (p.41). To establish an antitrust violation using the EFD, it is necessary to demonstrate that this refusal involves an essential input and cannot be reasonably justified by valid business reasons.

In the 1990s the term “essential” meant that the duplication must be unfeasible or unreasonable, signifying more than just an inconvenience or some economic disadvantage (OECD, 1996, p.8). During that period, the notion of essential facilities controlled by a monopolist was closely associated with tangible infrastructure like bridges, harbor terminals, and telecommunication networks, or with bundles of rights acquired through government concessions, such as airport slots, interurban telecommunication services, or authorization to employ wire circuits (local loops) within urban areas.

Legal decisions in the US and EU have established a checklist to assess the legality of potential anticompetitive behavior under the EFD, resembling a legal standard. Guggenberger (2021b) references the well-known MCI Communications case test for ‘access request’, which includes the following criteria: “(1) control of the essential facility by a monopolist; (2) a competitor's inability practically or reasonably to duplicate the essential facility; (3) the denial of the use of the facility to a competitor; and (4) the feasibility of providing the facility.” (p. 308).9 Decisions by the European courts, most notably associated with the Bronner case, as highlighted by Dunne (2020), led the DGComp to compile a set of steps for analysis (in refusal to deal cases) in their Guidance paper (EU, 2009):

“[a] the refusal relates to a product or service that is objectively necessary to be able to compete effectively on a downstream market; [b] the refusal is likely to lead to the elimination of effective competition on the downstream market, and; [c] the refusal is likely to lead to consumer harm.”

Both checklists emphasize the criteria of ‘essentiality’, ‘necessity’ or ‘indispensability’ as the first step. The EFD in the US is much stricter, requiring the input to be ‘essential’ and held by a monopolist. Interestingly, in the US the ‘essentiality’ may be attributed to a requirement to use, while in the EU there is a broader description of indispensability. Therefore, as Dunne (2020) shows, indispensability is context-specific and largely depends on the effects of supply restrictions. In other words, there are no “effective substitutes, no prospect of self-supply” (p.80) in the European analysis. Indispensability can be assessed by evaluating whether the absence of the input hinders the ability to exert effective competitive constraint. This perspective eschews simplistic polar analysis of “practical convenience” on one side and “nor absolute necessity’” (p.80) on the other.

The EFD emerged within the context of the US private and unregulated environment. Even in this context, the doctrine had multiple interpretation, as well noted by Sally Van Siclen, from the OECD Secretariat: “each [meaning] having to do with mandating access to something by those who do not otherwise get access” (OECD, 1996, p.13). The Brazilian economic environment differs significantly from that of the US and shares more similarities with regulated markets in Europe. In this context, the EFD is more likely to arise when a firm holds a monopoly due to regulatory reasons or when a firm is publicly owned or operates in a market of public interest subject to regulatory rules. In such a context, the desired policy objectives can be attained by employing economic regulation as an alternative to relying solely on antitrust decisions.

Therefore, in Brazil, the EFD has been applied in cases associated with monopolies resulting from control of physical facilities, as well as those derived from monopoly of rights. This latter case, sometimes, is similar to the concept of a gatekeeper. The EFD relies on the analysis of the economic harm due to the control of a non-replicable asset (for economic, technical of physical reasons). The next section will explore the relationship between the theory of harm and the EFD doctrine. 

The theory of harm underlying the EFD relies on the hypothesis of leveraging market power between two vertically related markets, with the presence of a dominant company in one of these markets (typically upstream) and competitive rivals in the other (downstream). By controlling essential facilities that act as a bottleneck for companies operating in the competitive market, the dominant company can limit access to essential inputs and restrict the competitive capacity of its rivals, thereby establishing a dominant position for itself in the downstream market. In other words, the assumption is that the company holding control over the bottleneck has an incentive to extend (or leverage), through anticompetitive practices, the market power derived from exclusive access to the bottleneck to related markets that depend on it (LAFFONT and TIROLE, 2001, p. 97-8).

However, since the 1970s, the leverage hypothesis, which is also used to characterize the anticompetitive effects of tying practices, has faced strong criticism. The basic problem with this hypothesis was that if a company effectively holds a monopoly over an essential input, the full extraction of available monopoly profit can be achieved directly through the pricing of that input, without requiring any exclusionary practice to extend that market power to other markets.

Many antitrust authorities have published vertical conduct guidelines, which offer guidance for the analysis of abuse of dominance and provide a unified logical analytical structure for evaluating antitrust harm of various exclusionary conducts, such as vertical restraints. This approach was nicely presented by the International Competition Network -ICN (2018): In the case of vertical restrains the defendant must be capable of generating anticompetitive harm and must have the incentive to do so.10

The capability to generate harm from a refusal to deal stems from the absence of alternatives and/or the significance of the refused input to the downstream user, which arises from its indispensability. In cases where an input is totally controlled by a monopolist, the lack of alternatives is typically evident immediately. In ordinary cases, the indispensability of the input encompasses both the lack of alternatives and the impact on the costs of the input users.

The (lack of) incentives to generate harm from leveraging market power from one market to another through a refusal to deal has been addressed by the single monopoly profit hypothesis (SMP) in the analysis of essential facilities cases. The argument of SMP can be presented in the case of a potential tying of two products, where Product A is offered under monopoly conditions and Product B under competitive conditions (BORK, 1978, p. 228-31).11 If the products are sold separately, the price of A (p_a) represents the monopoly price, and the price of B (p_b) a competitive price, thus equal to its marginal cost cb. However, if the products are bundled together at no additional cost, and v is the maximum price consumers are willing to pay for the AB system, the monopolist's profit per unit bundle from offering A will be v - c_b - c_a. Therefore, if the monopolist engages in tying, becoming the sole provider of B as well, their profit per unit will necessarily be the same: v - c_b - c_a.12 In other words, market power leverage does not create additional monopoly profit to be extracted. Thus, the practice of tying, bundling, or refusal to deal could only be explained by efficiency reasons (efficiency gains in joint production and distribution, contractual efficiencies).

Applied to the essential facilities doctrine, this argument suggests that: (i) there is no antitrust harm from restrictive practices of essential facilities holders with respect to trading with other firms; and that (ii) what is at stake is the exploitation of a pre-existing monopolistic position, not its leverage into additional markets and consumers. The latter issue should be resolved by regulation (and/or a regulatory agency) rather by the antitrust law (and/or an antitrust authority).

Nonetheless, as demonstrated in the post-Chicago literature, the SMP hypothesis does not hold universally. Its validity relies on the existence of specific conditions, which can often be violated. In the 1980s, the approach of Raising Rivals' Costs (SALOP and SCHEFFMAN, 1983; KRATTENMAKER and SALOP, 1986) pointed out that a conduct capable of limiting rivals' access to inputs or resources on reasonable terms can indeed generate significant anticompetitive effects. Such conduct would be both profitable and detrimental to consumers, involving the use of accumulated market power in one stage of the production chain to restrict competition in another stage, thus reintroducing an element present in the notion of market power leverage: the extension of market power to additional economic spaces.

But, as shown by Krattenmaker and Salop (1986), for potentially exclusionary conduct to negatively affect consumer welfare and be characterized as generating anticompetitive harm, two conditions must be met: (i) the conduct must effectively create competitive disadvantages (cost increases), which would depend on the characteristics of the input market involved; and (ii) the cost increases imposed on rivals (creation of competitive disadvantages) must enable the exercise of market power (price increases). The realization of anticompetitive harm would depend on both the presence of additional factors, such as the magnitude of cost increases for rivals (e.g., when there are economies of scale, as demonstrated by Whinston, 1990), resulting from the actual increase in input prices, and the existence of competition from rivals and/or contestable new entrants that are not subject to cost increases.

From the 1990s onwards, new research has analyzed the anticompetitive harm of exclusionary conduct by investigating the theoretical implications of contractual externalities. The identification of such contractual externalities has allowed for a better understanding of how a company controlling a bottleneck can reduce the competitive capacity of its non-integrated rivals even in transactions where they do not depend on the access being denied. To make this happen, it is sufficient that, as a result of economies of scale or network externalities, the loss of sales in transactions that depend on access to the bottleneck implies an increase in costs that affects the competitive capacity of the non-integrated company across all its customers. That is, including sales that do not directly depend on access to the bottleneck (WHINSTON, 2001; BERNHEIM and HEEB, 2015).

In this way, the exclusionary practice enables the extraction of additional monopolistic profits beyond those obtained through the pricing of the essential input, as would be the case in the theory of single monopoly profit. There is an economic mechanism that permits formulating the leverage hypothesis and explaining how a monopolist can restrain access to a product or input to extend its market power to adjacent markets. In the case of tying, if the monopolist of product A ties its sale to product B, the presence of economies of scale in the production of B can negatively affect the competitiveness of B producers even in transactions with consumers who only acquire product B. In other words, the exclusionary practice in the sales of the combination AB allows the monopolist to affect rivals and extend its market power to standalone sales of B.13 Similarly, in the case of an essential facility, the company that controls the bottleneck can deny access to rivals and thus extend its monopoly power to transactions or segments that do not depend on this access.

From the discussion, it appears that the EFD is a special case in the collection of antitrust analysis regarding refusal to deal and antitrust harm generated by leveraging market power from one market to another, by a dominant firm the in the former market. The table below presents the differences and similarities in the analysis, taking into account both the ability and incentives to exclude competitors or abuse market power. 

Digital platforms are ubiquitous in our daily lives, playing a significant role in communication, entertainment, trade and business operations. This topic has attracted significant interest in the past decade with the publication of numerous reports and papers.14 Platforms are businesses that facilitate transactions between buyers and sellers in an environment with significant cross-network externalities.15 They operate in two-sided and multisided markets, often referring to themselves as two-sided platforms. The sides are often individual consumers on one side, and business on the other, both platform users and purchasers of platforms services.

Digital platforms share characteristics with other digital businesses, such as a cost structure with high fixed costs and low or zero incremental costs, as well as economies of scale and scope in their cost and technical structure. However, they differ from other digital business as cross network externalities are key for their profitability and business model. Their business is data intensive, as other digital services. They also use algorithms to personalize the offering of goods, reducing search costs and enhancing user satisfaction with the platform. Two-sided platforms often dominate markets where innovation creates significant competitive advantage, with competition for the market.16

While platforms, due to their cost structure and scalability, may face high levels of entry, the network externalities and, arguably, services customization create significant barriers to entry and may lead to a winner-takes-all or winner-takes-most outcomes. Consumers often single-home, i.e., consume from only one provider, facing switching costs, whether direct or behavioral. Consequently, platforms become gateways to reach large groups of consumers within a specific market, exercising immediate control, i.e., gatekeeper control.17

Platforms also create their own eco-systems, either through providing a technical environment for other firms to operate (such as hardware operating systems) or by stablishing unique rules of trade (institutions) that require many firms to abide by their rules and practices in order to reach consumers on the other side(s) of the platform.

The ‘essentiality’ of platforms is often associated to situations in which trading with the platform is the way to reach consumers in the other side of the platform, acting thus as a ‘gatekeeper’ or ‘bottleneck’, including but not limited to: (i) very large market shares on one-side of a two-sided market organized by a platform; (ii) data availability required to design and efficiently operate the provision of services and goods to consumers in the other side of the platform (CREMER et al. 2019); (iii) technical standards or patents that must be used by sellers to market products in the (eco-system) platform (GUGGENBERGER, 2021b).

The dynamics of digital platforms exploit two concepts: bottlenecks and gatekeepers. Bottlenecks, as earlier defined by Jacobides et al (2006), refer to scarce and critical resources that improve the performance of a system. Firms that control platforms can benefit from innovation by managing industry bottlenecks with the goal of being the leader within the market. Firms that control value-creating assets to be traded by digital platforms aim to control a set of crucial gatekeeper functions, i.e., they intend to command the bottlenecks and manage access to a particular industry. When platforms achieve a position of dominance over access to a large group of users, they are called gatekeepers (BALLON and EVENS, 2014).

The gatekeeping position of a platform may give it an advantage in generating and collecting an input deemed indispensable for effective competition in digital markets, namely, data. Due to their extensive user base, gatekeeper platforms collect much more data than their competitors, giving them a considerable informational advantage (NICOLETTI et al, 2023).

Platforms that control gatekeeper functions can achieve higher margins in some environments by positioning themselves as mandatory bottlenecks between platform business users and customers. Market power in digital platforms can be exercised through the control of a network bottleneck. Bottleneck power arises from both supply-side conditions (control over essential facility or input required by competing producers) and demand-side conditions (consumers’ propensity to use only one platform for specific functionalities) (LIANOS and SMICHOWSKI, 2021).

As a gatekeeper, the platform company has the incentive to expand into new markets by providing a wide range of products and services, thereby establishing a stronghold over its customers within its product ecosystem. This strategy enhances the gatekeeper power of the company. It effectively becomes the arbiter of access to its users, exerting control not only over the entry of third parties into its user base but also over the access of its users to content, products, and/or services provided by third parties (BOURREAU and DE STREEL, 2019; OECD, 2023). Other possible examples of strategies to enhance market strategy are the use of exclusive contracts, bundling, or technical incompatibilities (STIGLER CENTER, 2019).

When platforms are both gatekeepers and competitors in a market, such a position allows the platform to assume the role of “private regulators.” Therefore, they can determine the participation rules in the exchange of goods and services (SCHWEITZER et al., 2019). This includes monitoring contracts concluded through the platform, establishing community guidelines, and implementing a reputation system that works as a governance architecture. In doing so, platforms often rely on algorithm-controlled matching systems that guide interactions between supply and demand, as well as among platform users and citizens (BUSCH et al., 2021).

For instance, according to the Competition and Market Authority (CMA) (2019), the UK’s Competition Authority, although firms are not required to act neutrally or provide access to their competitors, dominant platforms can prioritize the sale of their own products/services. This practice can be detrimental as it raises barriers to entry and expansion for competitors, while potentially diminishing quality and innovation in different markets.

The gatekeeper power and digital market characteristics, such as cross-network externalities, enable the actual exclusionary power of denying access to platform consumers. This exclusionary power can be achieved even if the platform is not a monopolist. Firstly, the platform must be sufficiently large that the alternative options available to the platform users cannot reach turning points in cross-network externalities. Secondly, network effects act as entry barriers, allowing the refusal to deal potentially block efficient competitors from entering and stifling innovation (GUGGENBERGER, 2021). Lastly, since the market sides are often prone to concentration, the monetizable sides may lack competition, thereby creating an opportunity to reap monopoly profits by controlling them.

The debate surrounding essential facilities and gatekeepers related to digital platforms can be exemplified by examining the assessment of Google Shopping antitrust cases, which has faced different outcomes in various jurisdictions. While the European Commission imposed significant sanctions on the company, the United States Department of Justice (DoJ) and Brazil's competition authority (CADE) also initiated investigations, deciding not to impose restrictions in their final decisions. Our goal here is not to review in detail the analysis for the decision in Brazil or compare the decisions across jurisdictions.18 Instead, we review the case pointing out that the analysis went beyond the EFD, even when concluding that the service provided by Google was not an ‘essential facility’.

In Brazil, E-Commerce Media Group, the holding company of price comparison services (PCs) Buscapé and Bondfaro, lodged a complaint against Google. They argued that Google would be favoring Google Shopping by displaying it in a privileged position in Google Search, and by scrapping content from competitors. From E-commerce’s point of view, Google acts as a gatekeeper for price comparison services since its search engine could be considered essential and could not be reasonably substituted, thus justifying its essentiality.

Both the General Superintendence (SG) and the Department of Economic Studies (DEE) examined Google's dominant position. They considered generic search engines and price comparison services as distinct and vertically integrated markets, with the latter being dependent on the former.19 The SG and DEE stated that that Google held a dominant position in the generic search engine market, but faced competition on the advertising side of the market. Nevertheless, this would not prevent Google from reducing the quality of Google Search, given its position on the users' side and the presence of barriers to entry, such as (i) network effects, (ii) data-related barriers, and (iii) users' inertial behavior.20

The Reporting Commissioner, Mauricio Maia, divided its investigation into six categories: (i) exclusionary practices, (ii) tying (iii) predatory innovation, (iv) privileged positioning, (v) lack of transparency and misleading advertising, and (vi) lack of transparency in its algorithms. The alleged practices were related to Google Search's Product Universal and Product Listing Ads (PLA) services. Product Universal, launched in the USA in 2007 and in Brazil in 2011, allowed for the comparison and ranking of generic and vertical search results. It enabled products to appear on the generic search results page with images, links, and prices. On the other hand, PLA, an advertising service introduced in the USA in 2010, displayed ads alongside generic search results, featuring images, seller names, and prices. In Brazil, Product Universal was discontinued in 2012 and replaced by PLA, which required sellers to pay a fee to have their products featured on the search results page.21 CADE examined anticompetitive practices related to both the Product Universal and PLA services.

When discussing exclusionary practices, commissioner Maia investigated whether Google Search could be considered an essential facility for price comparison services and therefore, unlawfully blocking access to it. As mentioned earlier, for the Essential Facility Doctrine (EFD) to apply, the duplication of the facility must be unfeasible or unreasonable. Commissioner Maia stated that the investigated company must: (i) control access to the essential facility, thereby monopolizing it (or having close to a monopoly) for a long-term period; and (ii) such control must hinder vertically related firms from competing effectively.

Commissioner Maia concluded that Google does not hold an essential facility, despite its market share. The commissioner investigated whether three features could be considered essential facilities: ads with pictures and sub-links; the first page of search results; and data. It was concluded that they were not. His decision rested on the following arguments. Firstly, when considering advertising as a distinct market, separate from price comparison services, there is substitutability without vertical integration. Secondly, the first page of search results cannot be an essential facility due to limited display capacity and personalized variations and consumers have direct access to price comparison websites and alternative search engines. Finally, while Google possesses significant user data and there are data-related barriers, the reporting commissioner concluded that price comparison services can function effectively without personal data and as it is non-rival and non-exclusive, is readily available from various sources and quickly becomes outdated. Based on Commissioner Maia's interpretations, none of the three services were deemed essential facilities, and therefore, the EFD would not be applicable.

The case proceeded to investigate the effects on competition, despite not falling under the Essential Facilities Doctrine (EFD). Commissioner Maia examined two exclusionary practices: refusal to deal and discrimination, along with other mentioned practices. In summary, the debate focused on Google's ability to leverage its dominant position to promote Google Shopping. Maia investigated two perspectives: vertical foreclosure and tying. The first perspective investigated whether the company would potentially exploit its position in a vertically integrated market to favor its own product and hinder competitors' access. The second perspective involved tying Google Search and Google Shopping.

After assessing the effects of these practices, Maia did not consider them anticompetitive and voted to dismiss the investigation. Analyzing all the alleged practices, he argued that: (i) the presence of price comparison services on the first page of Google Search results did not decrease after the introduction of Product Universal and PLAs (which differs from similar investigations in France and Spain), indicating no intentional manipulation of algorithms; (ii) there was no decrease in traffic from Google Search to price comparison services; (iii) there was no raise in rivals' costs through a rise in cost per click; (iv) Google Shopping charging for ads was not considered anticompetitive, as other E-Commerce platforms followed a similar practice; (v) restricting the number of results in Google Shopping was not deemed anticompetitive, as it aligns with Google's ad-based business model; (vi) the reduction in the number of players in the price comparison services market from 5 to 3 during the investigation was not attributed to Google's actions.

Although the decision was far from unanimous, CADE's Administrative Tribunal ultimately sided with Commissioner Maia's position. The Tribunal President cast the deciding vote to settle the decision. The commissioners who held different opinions from Commissioner Maia were Paula Farani Silveira, Paulo Burnier and João Paulo Resende.

Commissioner Burnier stated that the potential negative effects were sufficient  to consider the  practice as anticompetitive, particularly in digital markets. Consequently, he proposed reversing the burden of proof. He also stated that, in this specific case, Google failed to substantiate how its alleged innovation efficiencies would counterbalance the potential anticompetitive effects.22

Commissioner Resende contended that Google’s substantial market share in one market (general search) was tipping the competition in the price comparison search market. Thus, he considered it an antitrust violation. While Google was not a monopolist in the general search market, its behavior in this market was affecting consumers in other markets. In a sense, Commissioner Resende’s view is that the ‘essentiality’ does not require monopolization to generate effects.23

Commissioner Silveira considered that Google's conduct was anticompetitive and constituted an antitrust violation. In her argumentation, she emphasized that Google maintained a clear dominant position, supported by network externalities, lock-in effects, and economies of scale and scope. This dominant position was being leveraged to gain market power, discriminate against competitors, and cause harm to consumers. Thus, she disagreed with Commissioner Maia regarding the identification and substantiation of competitive harm. It is important to note that Commissioner Silveira explicitly highlighted that the existence of competitive harm does not require the presence of essential facilities. Her report serves as an illustrative example of how a theory of harm applicable to digital platform markets can be outlined without resorting to the essential facilities doctrine.24

As discussed by Fernandes (2021, pp. 303-312), the divergence between the commissioners reflected a dilemma in the application of antitrust policies in the repression of unilateral conditions in the digital platform market. The application of the EFD offered a more direct way to consider that practices such as self-preferencing should be considered anticompetitive, unless the company that implements it demonstrates an efficiency justification and the absence of anticompetitive effects. However, if the platforms are not considered essential facilities, two options arise: (i) using the theory of harm generally applied in cases of tying and raising rivals' costs, which places the obligation on the antitrust authority to demonstrate the anticompetitive effect and consumer harm; or (ii) consider that the internet gatekeeper position of some platforms justifies the application of a different theory of harm, in which the burden of proving the absence of anticompetitive effects is allocated to the defendant. The challenge for this second alternative lies in specifying more clearly the underlying economic reasoning of this alternative theory of harm, differentiating it from EFD and exempting it from the need to assume a form of ‘essentiality’. This is an open topic, to be clarified with the advancement of research and debates in the area.